As we were working through the equations, we noticed something unexpected. When we entered the same basic bearing and operating information into the different bearing manufacturers’ equations, they produced different answers. In some cases, the recommended grease quantities differed. In others, the calculated relubrication intervals varied considerably. These were not minor differences created by rounding or arithmetic mistakes. The equations themselves were producing different recommendations for what was supposedly the same lubrication problem.

The student understandably asked which answer was correct.

I could explain how to perform the calculations. I could explain the meaning of the variables. I could discuss bearing geometry, speed, temperature, loading, contamination and lubricant selection. What I could not adequately explain was why equations from respected bearing manufacturers produced different answers when applied to the same example.

Why bearing lubrication equations produce different answers

The following year, a colleague and I presented a paper at the National Lubricating Grease Institute’s annual conference titled How Much Grease Should This Bearing Receive and How Often Should It Be Lubricated? We shared the equations, worked through the mathematics, and presented the resulting discrepancies to an audience of more than 300 people representing lubricant manufacturers, grease formulators, equipment suppliers, researchers, engineers, and lubrication professionals.

I expected disagreement. I expected someone to challenge our assumptions, identify a variable we had overlooked, or explain that the equations had been developed for different bearing populations, operating environments, grease properties, or risk tolerances.

What I did not expect was silence.

Among an audience containing some of the most knowledgeable grease professionals in the industry, no one offered even a weak technical explanation for why the equations produced different results.

That experience stayed with me because it exposed something deeper than a disagreement among formulas. If the accepted equations do not produce a common answer, then the problem cannot be reduced to selecting the correct formula and performing the algebra properly. Each equation must contain assumptions, simplifications, empirical relationships, and safety factors that are not completely visible to the person using it. The equations may look precise, but the systems they attempt to describe are not.

How much grease does a bearing really need?

So why is determining the proper lubrication quantity and frequency still so difficult?

The most common question I am asked: How much grease should this bearing receive, and how often should it be lubricated? I always answer the same; it depends.

Go online and chances are the answer will involve charts, bearing dimensions, operating hours, or perhaps a reference to ultrasound or vibration analysis. Over the past several decades, our industry has invested heavily in technologies intended to remove uncertainty from lubrication practices. We have ultrasonic detectors that can hear friction before humans can. We have vibration analyzers capable of detecting microscopic defects inside rolling element bearings. We have infrared cameras, automatic lubricators, grease guns with calibrated outputs, sophisticated lubricants, and software that schedules every lubrication activity. Yet despite all of this technology, overgreasing and undergreasing remain persistent causes of premature bearing failure.

Why?

The obvious answer is to assume that technicians need better training or better procedures. While those certainly help, they fail to explain why experienced organizations equipped with capable people and advanced technology continue to struggle with the same problem. The deeper reason is that we have been trying to solve the wrong kind of problem. We have treated bearing lubrication as though it were a measurement problem when, in reality, it is an inference problem.

Grease is an oil reservoir, not a fuel tank

Most people imagine grease as something that is gradually consumed until it must be replenished, much like fuel in a tank. A bearing simply needs more grease when its existing supply becomes low. Unfortunately, bearings do not operate that way. Grease is not merely a lubricant placed between two moving surfaces. It is a semisolid system that holds base oil within a thickener structure and releases that oil over time. In most rolling element applications, the grease functions primarily as an oil reservoir, while also helping exclude contamination, control leakage, and maintain lubricant near the contact zone.

The rolling elements are therefore not riding on a large mass of grease. They are separated from the raceways by an extremely thin lubricating film whose formation depends upon the base oil, viscosity, speed, load, temperature, surface condition, grease mobility, and the ability of the grease to release and replenish oil near the contact.

This immediately creates a difficult question. We are not merely interested in how much grease remains somewhere inside the housing. We need to know whether enough usable oil is reaching the rolling contacts at the proper rate and whether that condition can be sustained until the next lubrication event.

No handheld instrument measures that internal state directly. Every measurement we make is indirect.

Why lubrication measurements are only indirect

Ultrasound has become one of the industry’s preferred lubrication tools because increasing friction often produces greater high-frequency acoustic activity. As lubricant is introduced and friction decreases, the ultrasonic signal may decline. The process appears almost elegant: measure the signal, add grease gradually, observe the response, and stop when the signal reaches an acceptable level.

Reality is considerably more complicated because lubrication condition is only one source of ultrasonic energy. Contamination, corrosion, electrical fluting, false brinelling, misalignment, excessive preload, insufficient internal clearance, damaged raceways, and other mechanical defects can also increase ultrasonic activity. Even the method used to collect the reading, including sensor placement, contact pressure, background noise, machine loading, and operating speed, can influence the result.

The technician standing in front of the bearing is therefore forced to answer a question the instrument cannot answer by itself: Is this bearing asking for grease, or is it telling me that something else is wrong?

Vibration analysis introduces another perspective. Modern vibration systems can identify remarkable details about machine and bearing condition. They can detect imbalance, looseness, misalignment, resonance, spalling, and rolling element defects long before catastrophic failure occurs. However, vibration analysis frequently becomes most valuable after physical distress has begun developing. Lubrication problems often begin at a much smaller scale through changes in film thickness, friction, asperity interaction, localized heating, and surface distress. By the time these conditions create an unmistakable vibration signature, the bearing may already be accumulating irreversible damage.

Vibration analysis is extraordinarily powerful, but it frequently observes the consequences of inadequate lubrication more clearly than it observes the earliest transition into lubricant starvation.

Why identical bearings can require different lubrication intervals

The problem becomes even more complicated because no two bearings live identical lives, even when they appear identical on paper. Consider two electric motors installed side by side during the same outage. Both contain the same bearing designation, use the same grease, operate at the same nominal speed, and carry nearly identical loads. One may require lubrication every three months while the other performs satisfactorily for nearly a year.

The difference is not necessarily luck or inconsistency. Small differences in alignment, shaft loading, bearing fit, internal clearance, housing geometry, seal condition, contamination exposure, operating temperature, vibration, mounting stress, and previous maintenance history continuously influence how grease moves, ages, releases oil, and reaches the rolling contacts. Every bearing develops its own operating history, and that history gradually becomes part of its lubrication requirement.

This also helps explain why the bearing manufacturer equations can disagree. The equations are not direct expressions of an immutable physical law. They are models developed from particular assumptions, test populations, empirical observations, application experience, correction factors, and acceptable levels of risk. One manufacturer may place greater emphasis on bearing geometry. Another may apply different speed factors, temperature corrections, contamination assumptions, or grease-life relationships. Some equations may be intentionally conservative because the consequences of insufficient lubrication are considered more serious than the consequences of additional maintenance. Others may assume operating conditions that are rarely made explicit when the equation is reproduced in a training manual or maintenance procedure.

The mathematics may be exact, but the assumptions beneath the mathematics are not universal. An equation can therefore produce a precise answer without producing the uniquely correct answer. This is one of the most dangerous characteristics of engineering calculations: the number of decimal places can conceal the uncertainty of the model.

Even the grease itself refuses to behave in a perfectly predictable manner. Unlike a material that is consumed at a constant rate, grease can undergo nonlinear aging. For a period of time, the thickener structure may release oil as intended and remain mechanically stable. Then oxidation accelerates, contamination accumulates, mechanical working changes the thickener structure, or elevated temperature rapidly consumes antioxidant protection. Oil separation may become excessive, or the grease may harden and lose its ability to replenish the contact.

What appeared to be a healthy lubricant during one inspection may deteriorate much more rapidly before the next. The transition from adequate lubrication to marginal lubrication does not necessarily follow the orderly schedule established in the computerized maintenance management system.

Where does the grease actually go?

One of the greatest unknowns remains something surprisingly simple: where does the grease actually go after it leaves the grease gun?

Three strokes of a grease gun sound precise until we realize how little that number tells us. Different grease guns discharge different volumes per stroke. Some of the grease may remain within the fitting, extension line, or lubrication passage. Some may enter the housing but remain away from the rolling elements. Some may pack against a seal. Some may purge immediately through a relief path. Some may be pushed into the bearing cavity and begin churning, increasing operating temperature without improving the lubricant film at the contact surfaces.

The quantity discharged from the grease gun is therefore not necessarily the quantity delivered to the bearing, and the quantity delivered to the bearing is not necessarily the quantity available to lubricate the contact. Without physically observing the internal distribution of grease, we are once again forced to infer what probably occurred rather than directly measure it.

Then there is the greatest source of variation of all: people.

Different technicians apply different numbers of grease gun strokes. Different grease guns deliver different quantities. Some technicians grease until they see lubricant purge from the housing. Others stop after a prescribed number of pumps. Some inject grease quickly, creating pressure and potentially forcing grease toward seals or shields. Others add grease slowly enough for the bearing to redistribute it while operating. Some technicians listen with ultrasound throughout the process. Others treat lubrication as a scheduled task that must be completed whether the bearing demonstrates a need or not.

Every one of these differences influences the resulting lubrication state, often more significantly than the apparent precision of the equation used to establish the work order.

Why bearing lubrication requires multiple sources of evidence

When all of these uncertainties are considered together, an important realization emerges. Ultrasound, vibration, temperature, motor current, operating history, grease chemistry, environmental conditions, bearing design, and maintenance records are not competing sources of information. They are individual pieces of evidence describing an internal condition that no single instrument or equation can observe directly.

This changes the role of the lubrication professional and reliability engineer. The objective is no longer to find the one formula or measurement that reveals the answer. The objective is to assemble multiple imperfect observations into the most accurate and defensible understanding of bearing condition that the available evidence allows.

The manufacturer’s equation provides an initial estimate. Bearing geometry and speed establish a starting point. Temperature, contamination, and loading modify that estimate. Ultrasound may indicate a change in frictional behavior. Vibration may reveal developing mechanical distress. Grease consumption history, purge condition, operating response, and previous outcomes provide additional context. Each observation increases or decreases our confidence, but none eliminates uncertainty completely.

Lubrication therefore becomes an exercise in disciplined reasoning under uncertainty rather than simple measurement.

Perhaps this explains why so many lubrication programs eventually plateau. Organizations become exceptionally good at collecting data while remaining only marginally better at making lubrication decisions. They purchase ultrasound instruments, collect vibration spectra, establish routes, calculate grease quantities, and enter frequencies into their maintenance systems. Yet unless the information is integrated, challenged, and continuously compared against actual bearing outcomes, the organization has only automated its assumptions.

Data alone rarely eliminates uncertainty. It provides additional evidence that must be interpreted within the broader operating context.

A mature lubrication program should therefore treat the calculated quantity and interval as hypotheses rather than permanent truths. The initial recommendation should be applied carefully, observed, and adjusted using evidence from the machine. Did the bearing temperature rise after lubrication? Did the ultrasonic signal decline and remain stable? Was excessive grease purged? Did the grease appear oxidized, contaminated, hardened, or separated? Did the bearing reach its expected service life? Did similar machines respond in the same manner?

Each lubrication event becomes a small experiment. The organization acts, observes the consequences, learns, and modifies the recommendation. Over time, the program moves away from generic intervals and toward machine-specific knowledge.

The future of evidence-based lubrication

The future of lubrication will almost certainly not depend upon a single revolutionary sensor or a universally accepted equation. It will depend upon intelligently integrating many sources of evidence into continuously improving estimates of bearing condition. Artificial intelligence, physics-based models, historical maintenance records, operating context, lubricant chemistry, ultrasound, vibration, temperature, and reliability history may eventually work together to estimate something that today remains largely invisible: the true lubrication state of a bearing.

That represents an important philosophical shift for our profession. The challenge has never been simply determining how much grease to apply or how frequently to apply it. The real challenge has always been learning how to make better decisions when direct observation is impossible and certainty is unavailable.

The equations that my student and I worked through were not necessarily wrong. Each was an attempt to approximate a complicated physical system through a manageable mathematical model. The mistake would have been assuming that because the equations produced numbers, one of those numbers had to represent a universal truth.

Perhaps the question we should have been asking all along was never simply, How much grease does this bearing need?

Or maybe the better question is, Given the model, the machine, its operating history, and everything we can observe today, what is the most defensible lubrication decision we can make?

That is the essence of reliability, the disciplined management of uncertainty.